The present invention relates to an apparatus for using a control policy to implement power saving features.
Large computing systems are very power-intensive, consuming electricity in huge quantities as disks, servers etc. remain powered on. This is because customers often want a high-availability, high-performance and high-reliability service.
It is desirable for owners of these systems to conserve power, thereby minimising the impact on energy sources and also minimising running costs.
In the prior art, an algorithm for power management can be executed in a computing system. The algorithm places frequently used data on a subset of available disks, such that the remaining disks hold less frequently used data. The algorithm achieves power savings because the disks that hold less frequently used data can be operated in power saving modes. For example, these disks can be spun down and only spun up occasionally in order to service an infrequent data access.
Disadvantageously, by concentrating frequently used data on a subset of available disks, the useable I/O processing capacity of the computing system is reduced. As the I/O workload increases, the disks that hold frequently used data will reach the limit of their I/O processing capacity. As this limit is approached, a user will experience increasingly longer I/O wait times and thus, performance will degrade.
A solution to this problem is to distribute frequently used data over several available disks in order to maximise the effective I/O processing capacity of the computing system.
Thus, the skilled person is faced with a conflict: in order to minimise power consumption, frequently used data should by concentrated on a minimum number of disks and in order to maximise performance, frequently used data should by distributed over a maximum amount of disks.
There is a need for an improved power management mechanism that can be implemented whilst minimising impact on customer service levels.